A thorough systematic review was conducted according to the Cochrane methodology for systematic reviews as described in the Cochrane Handbook for Systematic Reviews of Interventions [11] (Multimedia Appendix 1). Our trial was registered with PROSPERO (registration number CRD420261346230). A literature search was conducted in Scopus, Medline, CINAHL, PubMed, CENTRAL, and Google Scholar databases in April 2024 using relevant search terms associated with the research question, “Can wearable device data be used to effectively augment postoperative patient-reported outcomes after total joint replacement?” (Multimedia Appendix 2). Literature searches were not limited by date of publication.

A total of 471 papers were identified from the databases or registers used, as well as 7 studies from the reference lists of other papers. Selection criteria included availability in English and full-text accessibility. Systematic reviews and theoretical papers were excluded. After the removal of 100 duplicates, the abstracts of 371 studies were screened for compliance with the criteria. Further, 269 studies were excluded during abstract screening due to a lack of adherence to selection criteria, resulting in 102 studies to undergo full-text review. Finally, 35 studies were identified from the full-text review that satisfied the selection criteria for inclusion in this systematic review.

Data extraction of the 35 studies was performed on Covidence. Data extraction was originally performed independently by 2 reviewers, then reviewed for consensus. Extracted data included information regarding study identification, aims, study design and population, inclusion and exclusion criteria, data collection characteristics, interventions, accelerometer details and data, PROMs and quality of life assessments used, and study end points.

Study quality and bias risk were assessed using Cochrane’s Risk of Bias tool for randomized trials (RoB-2) [12] and the Risk of Bias in Nonrandomized Studies of Interventions (ROBINS-I) tool [13]. RoB-2 includes 5 domains: the randomization process, deviations from the intended intervention, missing outcomes data, measurement of the outcome, and selection of the reported result. ROBINS-I evaluates 7 domains: confounding, participant selection, intervention classification, deviations from intended interventions, missing data, outcome measurement, and selection of reported results. RoB-2 categorized bias as low risk, some concerns, high risk, or no information, while ROBINS-I classified risk as low, moderate, serious, critical, or no information. A total of 2 authors independently assessed the study quality of each paper and resolved any concerns with thorough discussion. Traffic light plots of domains regarding risk of bias were generated using the risk-of-bias visualization tool [14] (Multimedia Appendix 3). Notably, 5 studies were funded by Zimmer Biomet [15-19], in 1 study [20], 360 Knee Systems supplied the wearable device, and in 1 study [21], the lead author was a creator of the MoveUp Application that was used to assess patient recovery.

The quality of the studies was assessed using RoB-2 for the 5 randomized controlled trials (RCTs) and ROBINS-I for the 30 nonrandomized studies. RCT data are presented in Table 1, and nonrandomized data in Table 2. For overall risk of bias, 5 studies were characterized as low risk, 16 studies were ranked moderate risk or some concerns, 11 studies were ranked serious risk, and 3 studies were characterized as not enough information.

The literature identified for review demonstrated considerable variability in variables collected via wearable devices as well as outcomes reported. As such, a meta-analysis was not performed. Rather, a systematic review was performed according to the Synthesis Without Meta-Analysis in systematic reviews guideline [49]. Studies were grouped according to outcomes reported, as demonstrated in Table 3. The results, limitations, and data presentation from our synthesis are also presented in Table 3.

This systematic review analyzed 35 studies that used wearable devices to assist with patient monitoring in the TJA perioperative period. These papers detail how monitoring with wearables is used in conjunction with PROMs and functional metrics in the recovery of patients undergoing TJA. The final analysis included 35 papers, including 5 RCTs and 30 nonrandomized papers (Figure 1).

Of the 35 papers included in this study, 32 collected step count, making it the most common data point reported (Table 3). Another component of assessing functional outcomes in TJA is performance-based tests, such as the 6-minute walking test (6MWT), range-of-motion, or timed-up-and-go (TUG; Table 3). A total of 9 studies used both wearable device data and standardized performance-based functional outcomes in their assessments (Table 3).

Physical activity and PROMs used were not consistent among studies (Table 3). Joint-specific PROMs, including Knee Society Score, Knee Injury and Osteoarthritis Outcome Score (KOOS), Oxford Knee Score, Hip Disability and Osteoarthritis Outcome Score (HOOS), and Harris Hip Score, were used in 21 studies. Virtual Analog Scale (VAS) was used in 4 studies [15,19,34,41] and Numerical Rating Scale for pain was used in 4 studies [15,18,24,47]. EQ-5D scores were examined in 4 studies [15,16,21,23], and 3 used Western Ontario and McMaster Universities Arthritis Index (WOMAC) [32,39,43]. The Lower Extremity Functional Scale (LEFS) was used in 2 studies [48,51]. The Veterans Rand 12-Item Health Survey was used in 2 studies [37,38], and 1 used SF-36 (36-Item Short Form Health Survey) [30].

Sleep as recorded by wearable devices was reported in 7 of the included studies (Table 3). Gibian et al [19] recorded sleep duration and reported a negative correlation with VAS pain scores [35]. A total of 2 additional studies [17,20,24] also report sleep duration as recorded by wearable devices. Frimpong et al [31] use a validated algorithm to report sleep data as recorded by wearables. Karas et al [44] report “sleep efficacy,” which is the proportion of minutes asleep to the total time in bed. Finally, Fallahzadeh et al [43] and Vitale et al [41] focused on multiple components of sleep, with the former including overall sleep and sleep quality and the latter recording 9 components of sleep.

Interestingly, only 1 study used a novel application to assess patients [28]. This study demonstrated that step counts as recorded in the MoveUp application correlated with LEFS following TJA. However, the authors did not mention how this application can be used to create tailored rehabilitation protocols.

This systematic review found that most of the literature regarding the utility of wearable devices to monitor TJA outcomes includes step count. In addition to recording step count, these devices can record data regarding gait speed and asymmetry as well. Assessment of such metrics allows providers to evaluate gait comprehensively. Nevertheless, few studies used this capability of wearables [16,17,22,46]. Further, there is a large variability in how wearable device data is used in conjunction with other outcome measures in the current literature. Further, while ample studies include PROMs in addition to wearable data, many different PROMs are used. As demonstrated in Table 3, wearable devices have the capability of measuring multiple objective metrics. While this capability has the potential to allow holistic assessment of patient recovery, there is currently no consensus on which metrics are most valuable to record and assess in the perioperative setting. Further, the studies reviewed are not uniform in when patients are assessed, with certain groups analyzing patients days to weeks after surgery, while others continue for 6 to 12 months. The heterogeneity among these studies demonstrates that the ideal manner in which surgeons can use wearable devices for remote monitoring requires further research. We suggest that stratification regarding the reporting of perioperative data obtained by wearables, and that coupling this data with a unified PROM, will maximize the utility of wearable devices as a tool for assessing patient recovery.

The relationship between wearable-derived physical activity and PROMs, such as pain, quality of life, and self-reported function scores, was explored by 21 studies. The most common PROMs reported were joint-specific. Multiple studies found a correlation between change in step count and KOOS, KOOS JR (KOOS-joint replacement), HOOS, or HOOS JR [15,18,20,36,38,42,47,48]. Both the KOOS and HOOS were developed particularly for patients with osteoarthritis and have long been used to track progress following TJA. Such studies support the use of wearable devices as another component of comprehensive monitoring of TJA recovery, particularly since such devices allow continuous monitoring. However, the current literature does not yet suggest a data point obtained via remote wearables that necessitates intervention or that intervention improves KOOS or HOOS.

Contrastingly, some groups found that steps did not correlate with other joint-specific PROMs. Frimpong et al 2020 [31] report that there is no correlation between KOOS and an increase in physical activity. Additionally, Tang et al [27] reported that there is no correlation between accelerometry-measured activity level and HOOS JR within 3 months of TJA. Multiple studies found no correlation between step count and WOMAC [31,43] or UCLA (University of California, Los Angeles Activity Scale) activity index [35,45]. It is unclear why many studies found that step count correlated with KOOS and HOOS, but not WOMAC or UCLA activity index. One suggestion is that HOOS and KOOS have been shown to better detect changes over time, while WOMAC and UCLA are more adapted for more active populations [52,53]. Regarding the UCLA activity index, it is possible to have large increases in step number but not have an increase in activity level on this scale.

In addition to joint-specific PROMs, other measures were used. Notably, 3 studies [28,43,48] found relationships between perioperative physical activity and patients’ ability to perform everyday tasks as measured by the LEFS and Surgical Recovery Scale. Others [20,36-38] found a positive correlation between wearable device reported step count and general health and physical functioning, as measured by the Veterans Rand 12-Item Health Survey PCS (Physical Composite Score) or SF-12 (12-Item Short Form Health Survey) PCS scores.

Further, some of the reviewed studies include functional outcome measures. Performance-based tests, such as the 6MWT and TUG, allow standardized assessment of patients. These tests are normally used immediately following surgery, before patient discharge, and then repeated at certain intervals in the clinic. As such, these tests allow serial, objective evaluation of patient recovery. Paxton found that both 6MWT and step count increased 20% from baseline at 12 weeks post-TKA, while TUG improvements are present but less robust [22]. Similarly, Pechstein et al [26] found a dose-response relationship between 6MWT and device-measured physical activity. Improvements in physical activity and TUG can be seen as early as 10 days postoperatively [41], and van Diemen et al [42] found a significant correlation between activity and TUG. Step count and 6MWT both measure endurance in walking, while TUG focuses on mobility. This correlation suggests that data collected by wearable devices might serve as a surrogate for these objective assessments. In contrast to TUG and 6MWT, wearable devices allow assessments to be performed continuously and remotely. As such, patients can be assessed in between clinic visits. When patients deviate from expected progress or demonstrate setbacks, tailored protocols can be developed.

While this review found that most wearable device literature includes step count, many wearable devices have the capacity to monitor additional variables, such as balance, as well. Further investigation could examine how such parameters correlate with TUG and other performance measures.

Further, because wearable devices are used during the preoperative period, they allow providers to obtain a patient baseline before TJA. The collection of this data allows recovery trajectories to be crafted. Patients demonstrate decreases in activity, balance, and gait speed acutely postoperatively, but then return to or surpass their baseline postoperatively. In the acute postoperative setting, knowledge of patient deviation from baseline can inform discharge protocols to allow patients to return to benchmarks. As patients progress in their recovery, comparison of metrics to preoperative and immediate postoperative data informs the multidisciplinary team about patient recovery. If a patient is not meeting a step count goal, therapy can be tailored toward endurance. Contrastingly, if a patient is struggling with balance, therapy can be adjusted to focus on improving this deficit. As the number of patients who have data collected increases, there is also potential to create expected recovery trajectories for patients based on baseline preoperative data, allowing providers to provide realistic and individualized recovery expectations.

In addition to papers that focus on activity, 7 of the included studies use wearable devices to monitor sleep in the TJA perioperative period. It has been well established that sleep and pain are closely associated, and it has been further shown that higher sleep quality is associated with better acute pain relief following TKA [28,29]. Wearable devices were shown to both increase physical activity on the Active Australia Survey and reduce time in bed and sleep duration [24]. This study did not report on sleep quality as reported by the wearable device. Another study found that both daily steps and minutes slept returned to preoperative levels 3 months post THA. Again, sleep quality as measured by wearable was not reported. Karas et al [44] report a “sleep efficacy score,” which is the proportion of minutes asleep to the total time in bed. They find that reductions in sleep efficiency from baseline up to 24 weeks postoperatively, but that daily steps return to baseline 12 weeks after surgery.

Gibian et al [19] found that patient-reported sleep quality, particularly poor-quality sleep as defined by the Pittsburgh Sleep Quality Index, correlated well with VAS pain score at 30 days postoperatively. However, they found no relationship between device-recorded sleep duration and VAS. Further, they found no correlation between overall Pittsburgh Sleep Quality Index scores and device-recorded hours of sleep at 30, 60, or 90 days postoperatively. Finally, 1 study used wearable devices to monitor multiple components of sleep, including overall sleep, sleep onset, and sleep disruption. Sleep disruption increased after surgery, but overall sleep did not. Sleep duration as an isolated metric may lack sensitivity in predicting changes in a patient’s recovery timeline following TJA, but fortunately, smartwatches and activity trackers can collect more advanced measurements of sleep. Taken together, the current literature that uses wearable devices to monitor sleep in TJA is heterogeneous. Different groups report an array of variables. This underscores the immense capacity of wearable devices to track multiple domains and also highlights the fact that more uniform outcomes reporting would allow studies to be more easily compared. Further, despite sleep data being collected and reported, current literature does not report on intervening on patients with altered sleep components as reported by wearables.

Notably, 5 of the 35 reviewed studies were sponsored by Zimmer Biomet [15-19], 1 study used a mobile application the lead author created [21], and 1 study used a wearable supplied by 360 Knee Systems [20]. Redfern et al [15] report that patients in the lowest preoperative step count group have lower KOOS JR scores and demonstrate greater postoperative improvement in KOOS JR. Similarly, Pasqualini et al [18] demonstrate that step counts correlate with HOOS JR both preoperatively and at 1- and 3-month follow-up. Further, 2 industry-funded studies [16,17] suggest that PROM surpasses minimal clinically important difference earlier in recovery than activity as measured by wearables. Twiggs et al [20] found a positive correlation between postoperative step count and KOOS activities of daily living subscore. Lebleu et al [21] perform an analysis using the MoveUp application the lead author created. They find that peak cadence improved beyond minimal clinically important difference more quickly than step count in both TJA and TKA recovery. Taken together, the industry-funded studies suggest that PROMs may improve faster than wearable device measurements and that some PROMs correlate with step counts or other measures of physical activity. Further, this research demonstrates the heterogeneity in the types of wearable data reported and the PROMs used among different studies.

The major strength of this study is the comprehensive examination of existing literature. Multiple databases were searched to ensure a comprehensive review of the literature. Further, 2 authors independently extracted data and then collaborated for consensus regarding study eligibility. There was also an agreement between reviewers regarding bias assessment according to the RoB-2 or ROBINS-I tools. This review reports on the vast capabilities of wearable devices, ranging from tracking step counts and sleep to more sophisticated measures. Moreover, we report how wearable device data correlates with validated PROMs and outcome measures, emphasizing their potential to revolutionize postoperative outcome monitoring. Wearable devices also encourage patient participation in rehabilitation protocols and offer patients objective data by which to monitor their postoperative progress. The studies included for review demonstrate high patient participation and compliance.

There are several limitations to this review. As with all reviews, the quality of the reviewed studies is a limitation. Our review included only 5 RCTs, with 30 studies representing nonrandomized trials, and these have lower levels of evidence. As discussed, the heterogeneity among study methods makes direct comparisons difficult. Different wearables have different capabilities. Further, different groups collected data at different time points or had patients wear devices in different locations. While the included studies provided patients with the wearable devices, requiring patients to use a wearable is a source of bias. Patients who are familiar with these devices may have higher technological literacy, socioeconomic status, or ability to care for themselves. As such, certain populations may be excluded from these studies. Additionally, the use of a wearable device on the wrist immediately postoperatively can result in falsely low recordings if patients are using an assistive device such as a walker or cane.

Finally, we propose that remote monitoring via wearable devices provides a practical method for tracking recovery, supporting patient engagement, and detecting deviations from expected progress. Current literature does not agree on the optimal data that wearables provide or when data collection should be performed. Despite this, the studies included in this review demonstrate that variables collected via wearable devices correlate with certain PROMs, pain measures, and objective postoperative tests. As such, monitoring patients with wearables may allow a more holistic assessment of patient recovery. Further, wearable devices have the potential to allow analysis informed by baseline preoperative activity, which can enhance preoperative counseling and personalize recovery expectations. Lastly, continuous monitoring may also allow early identification of delayed recovery, enabling timely intervention and tailored rehabilitation. Further studies can explore the ideal time in recovery, variables to assess, and the combination of data collected with additional outcome measures that allow wearable devices to most effectively assess outcomes following TJA.

The increasing prevalence of wearable devices presents significant opportunities for advancing postoperative care through continuous health monitoring and data collection. Remote monitoring using these devices allows surgeons to gather extensive and continuous data, which can be crucial for assessing outcomes following TJA. When integrated with a collection of PROMs and objective assessments, data from wearable devices facilitates more comprehensive patient surveillance and enables tailored interventions during the perioperative period. Early monitoring of patients in the postoperative phase enhances the timely identification of adverse events and complications, thereby allowing prompt intervention. Additionally, remote monitoring of patients has been demonstrated to improve compliance with postoperative protocols and instructions. Through rigorous evaluation and standardization of data collected, remote monitoring has the potential to elevate the quality of care and enhance the recovery experience for TJA patients, ultimately leading to improved patient outcomes and satisfaction. However, further research is needed to delineate which specific remote data points are most effective in facilitating the early identification of adverse events and in optimizing patient outcomes and satisfaction after TJA. Future studies can focus on which data provides the most valuable and complete assessment of patient recovery. As our knowledge of how wearables can be best used to assess outcomes increases, there is potential to use data provided to tailor continuous patient assessment and individualized rehabilitation protocols, customized postoperative evaluation, and early detection of deviation from norms in recovery.